Models / gpt-oss-20b

OpenAI

gpt-oss-20b

gpt-oss-20b trails most of the field on what it has been measured on; worth a look for simple, high-volume work where cost matters most. The cheapest model of similar intelligence is Llama 3.1 8B Instruct, at $0.06 per million tokens against $0.06.

Results 5 Nov 2026 · 38 results · 3 sources

Price per million tokens
$0.03 in · $0.15 out
Typical host on OpenRouter, 5 Nov 2026 · compare prices
Speed
277 tokens a second; on its test prompt the answer starts after 7.6 s, thinking time included
Artificial Analysis, on its usual API, at low reasoning; 263–277 across 2 settings
Intelligence Index
10.0
Artificial Analysis, at low reasoning; 9.0–10.0 across 2 settings
Spring Prompt overall
Not ranked yet
needs our benchmarks and two groups of results
Developer
OpenAI, based in the US
Weights published
Released
5 Aug 2025
131,072 tokens of context

Where it stands among the models you could choose

Available models grouped by price, intelligence, speed and developer location, plus models with published weights. Retired models are excluded.

AmongIntelligence IndexSpring Prompt overallPrice, blendedSpeed
Models available today 138th of 188Claude Opus 5.5 – $0.06 blended, 2nd percentilecheapest: Mistral Nemo 5th of 65Nemotron 3.5 Super VL
Models priced under $0.25 per million tokens 18th of 42Claude Haiku 5.5 – sets this group 2nd of 19Nemotron 3.5 Super VL
Models with similar intelligence sets this group – $0.06 blended, 3rd percentilecheapest: Llama 3.1 8B Instruct 1st of 14leads
Models writing 200 or more tokens a second 11th of 14Muse Spark 1.3 – $0.06 blended, 0th percentilethe cheapest sets this group
Models from developers based in the US 72nd of 93Claude Opus 5.5 – $0.06 blended, 2nd percentilecheapest: Gemma 3 4B 3rd of 33Nemotron 3.5 Super VL
Open-weights models 68th of 106MiMo-V2.6-Pro – $0.06 blended, 4th percentilecheapest: Mistral Nemo 4th of 43Nemotron 3.5 Super VL

The name under each rank is the leader of that group; green is the top quarter of the group and red the bottom quarter, by rank. Price is shown as a percentile: the share of the group that costs less, so lower is cheaper. "Sets this group" marks the measure the group is defined by. Spring Prompt overall is our score out of 100 across every source (how it works). Price is the developer's list price (or the typical OpenRouter host where we haven't read one) for three input tokens to one output; intelligence and speed are from Artificial Analysis (data sourced from Artificial Analysis); "similar intelligence" means within 4 points on its Intelligence Index. Developer locations are where each company is based, not where a model is served. Groups under 3 models are not ranked.

How good is it, and for what?

Each use case ranks the available models its benchmarks measured, then those priced under $0.25 per million tokens. The bar shows its percentile in that field, best to the right. Open a row for the results behind it.

Product listingsTurning a sparse product feed and photos into listings that can go live Not measured
  • CatalogBench: has not measured this model.
Decks from an analysisTurning a finished analysis into a deck you could present as it is Not measured
  • DeckBench: has not measured this model.
User surveysPlanning a user survey and reading its results without being misled Not measured
  • SurveyBench: has not measured this model.
Marketing planningPlanning a year of ad spend without overspending Not measured
  • ROASBench: has not measured this model.
Agents and tool useMulti-step tasks with tools: support desks, coding agents, function calls 129th of 160 available · 23rd of 38 at its price
Benchmarkgpt-oss-20bRank among availableBest available
Artificial AnalysisArtificial Analysis: Terminal-Bench 4.0 · reported 0.0% 65th of 106 Claude Sonnet 5.5 63.6%
Artificial AnalysisArtificial Analysis: Terminal-Bench 2.1 · reported 13.9% 86th of 108 Claude Fable 5.1 91.4%
Artificial AnalysisArtificial Analysis: τ-bench banking · reported 7.0% 83rd of 105 Grok 4.6 50.7%
  • tau2-bench: has not measured this model.
  • Berkeley Function Calling Leaderboard (BFCL) V4: has not measured this model.
  • OpenHands Index: has not measured this model.
  • Microsoft STATE-Bench: has not measured this model.
  • Vending-Bench 2: has not measured this model.
Professional workReal tasks from banking, consulting and law, business documents and freelance projects Not measured
  • APEX-Agents: has not measured this model.
  • GDP.pdf: has not measured this model.
  • Remote Labor Index: has not measured this model.
Reasoning and knowledgeHard questions across science, maths and general knowledge 133rd of 188 available · 17th of 42 at its price
Benchmarkgpt-oss-20bRank among availableBest available
Artificial AnalysisArtificial Analysis: Artificial Analysis Intelligence Index · reported 10.0 138th of 188 Claude Opus 5.5 57.6
Artificial AnalysisArtificial Analysis: Humanity's Last Exam · reported 11.0% 126th of 187 Claude Opus 5.5 61.4%
Artificial AnalysisArtificial Analysis: GPQA Diamond · reported 68.8% 120th of 171 GPT-6 Astra 96.3%
WritingWhat people prefer in blind comparisons, and judged writing quality 159th of 173 available · 31st of 40 at its price
Benchmarkgpt-oss-20bRank among availableBest available
Arena (formerly LMArena)Arena Text: overall · reported 1,318 136th of 156 Gemini 4 Argon 1,525
UGI LeaderboardUGI: writing score · reported 24.6 118th of 126 Gemini 3.8 Flash 78.6
Sticking to the factsSummarising without inventing things, and factual answers Not measured
  • Vectara Hallucination Leaderboard: has not measured this model.
  • Arena (formerly LMArena): has not measured this model.
  • SimpleQA Verified (Epoch AI): has not measured this model.
Speed under pressureGood decisions against a real clock (fast chess) Not measured
  • BulletBench: has not measured this model.

Nearest alternatives

Against gpt-oss-20b at $0.06 blended, Intelligence Index 10.0. "Similar intelligence" means within 4 points or better.

Where to run it

11 hosts on OpenRouter at the standard service tier, with a median of $0.06 blended per million tokens; quantisation where the host states it: FP8, FP4, BF16. Cheapest first.

HostQuantisationInputOutputBlended
DarkbloomFP8$0.02$0.09$0.04
AkashMLFP4$0.02$0.10$0.04
CoreWeaveFP4$0.03$0.13$0.06
DekaLLMBF16$0.03$0.14$0.06
DeepInfraBF16$0.03$0.14$0.06
ParasailFP4$0.03$0.15$0.06
PrimeIntellectFP4$0.04$0.17$0.07
SiliconFlowFP8$0.04$0.18$0.07

Prices per million tokens from OpenRouter's public endpoint list, checked 5 Nov 2026; the 3 dearer hosts are left out.

Against the alternatives

The models you would most likely weigh it against: the leaders of the groups above.

Scroll sideways to see every alternative.

gpt-oss-20bClaude Haiku 5.5
most intelligent at under $0.25 per million tokens
Llama 3.1 8B Instruct
cheapest with similar intelligence
Claude Opus 5.5
most intelligent available today
MiMo-V2.6-Pro
most intelligent with open weights
Price per million tokens $0.06 $0.20$0.06$8.00$0.54
Tokens a second 277 240–9643
Intelligence Index 10.0 43.46.957.646.3
Spring Prompt overall – 44–84–
Rank among available models, by use case
Product listings – 15th–10th–
Decks from an analysis – 14th–4th–
User surveys – 4th–3rd–
Marketing planning – 26th–7th–
Agents and tool use 129th of 160 23rd154th2nd19th
Professional work – ––3rd–
Reasoning and knowledge 133rd of 188 22nd166th1st12th
Writing 159th of 173 –168th5th22nd
Sticking to the facts – ––3rd10th
Speed under pressure – 19th–––

Shaded figures are better than gpt-oss-20b's. A dash means no result.

Where it has been measured

Fastest of the 14 models with similar intelligence, and cheapest of the 14 models writing 200 or more tokens a second. 18th most intelligent of the 42 models priced under $0.25 per million tokens. Among models available today, in the bottom quarter for agents and tool use, and writing.

Every result

38 published results from 3 sources, each in the source's own units, with the configuration that produced it.

Arena (formerly LMArena) · reported by Arena (formerly LMArena) · 6 results
MeasureValueRankConfigurationDated
Arena Text: business, management and financeArena rating, higher is better 1,323
1,309–1,337
239 of 406 gpt-oss-20b 2 Oct 2026
Arena Text: creative writingArena rating, higher is better 1,240
1,221–1,258
301 of 411 gpt-oss-20b 2 Oct 2026
Arena Text: expert promptsArena rating, higher is better 1,314
1,287–1,342
245 of 363 gpt-oss-20b 2 Oct 2026
Arena Text: instruction followingArena rating, higher is better 1,281
1,269–1,294
283 of 413 gpt-oss-20b 2 Oct 2026
Arena Text: overallArena rating, higher is better 1,318
1,311–1,324
263 of 413 gpt-oss-20b 2 Oct 2026
Arena Text: writing, literature and languageArena rating, higher is better 1,275
1,262–1,288
287 of 412 gpt-oss-20b 2 Oct 2026

Not shown: Not accuracy or correctness: it ranks which answer voters preferred, with answer length and formatting controlled for.

Artificial Analysis · reported by Artificial Analysis · 23 results
MeasureValueRankConfigurationDated
Artificial Analysis: Artificial Analysis Coding Indexindex score, higher is better 20.7 154 of 187 gpt-oss-20b (high reasoning) 20 Oct 2026
Artificial Analysis: Artificial Analysis Intelligence Indexindex score, higher is better 9.0 336 of 413 gpt-oss-20b (high reasoning) 20 Oct 2026
Artificial Analysis: Artificial Analysis Intelligence Indexindex score, higher is better 10.0 319 of 413 gpt-oss-20b (low reasoning) 20 Oct 2026
Artificial Analysis: GPQA Diamond% of questions, higher is better 68.8% 251 of 360 gpt-oss-20b (high reasoning) 20 Oct 2026
Artificial Analysis: GPQA Diamond% of questions, higher is better 61.1% 290 of 360 gpt-oss-20b (low reasoning) 20 Oct 2026
Artificial Analysis: Humanity's Last Exam% of questions, higher is better 11.0% 261 of 411 gpt-oss-20b (high reasoning) 20 Oct 2026
Artificial Analysis: Humanity's Last Exam% of questions, higher is better 5.3% 329 of 411 gpt-oss-20b (low reasoning) 20 Oct 2026
Artificial Analysis: IFBench% of instructions, higher is better 65.1% 88 of 288 gpt-oss-20b (high reasoning) 20 Oct 2026
Artificial Analysis: IFBench% of instructions, higher is better 57.8% 112 of 288 gpt-oss-20b (low reasoning) 20 Oct 2026
Artificial Analysis: Long-context reasoning (AA-LCR)% of questions, higher is better 34.7% 319 of 399 gpt-oss-20b (high reasoning) 20 Oct 2026
Artificial Analysis: Long-context reasoning (AA-LCR)% of questions, higher is better 31.0% 333 of 399 gpt-oss-20b (low reasoning) 20 Oct 2026
Artificial Analysis: Output speedtokens per second, higher is better 263 7 of 137 gpt-oss-20b (high reasoning) 20 Oct 2026
Artificial Analysis: Output speedtokens per second, higher is better 277 6 of 137 gpt-oss-20b (low reasoning) 20 Oct 2026
Artificial Analysis: SciCode% of problems, higher is better 38.9% 164 of 197 gpt-oss-20b (high reasoning) 20 Oct 2026
Artificial Analysis: Terminal-Bench 2.1% of tasks, higher is better 13.9% 151 of 188 gpt-oss-20b (high reasoning) 20 Oct 2026
Artificial Analysis: Terminal-Bench 4.0% of tasks, higher is better 0.0% 141 of 193 gpt-oss-20b (high reasoning) 20 Oct 2026
Artificial Analysis: Terminal-Bench Hard% of tasks, higher is better 10.6% 198 of 282 gpt-oss-20b (high reasoning) 20 Oct 2026
Artificial Analysis: Terminal-Bench Hard% of tasks, higher is better 4.5% 237 of 282 gpt-oss-20b (low reasoning) 20 Oct 2026
Artificial Analysis: Time to first answer tokenseconds, lower is better 8.01 53 of 137 gpt-oss-20b (high reasoning) 20 Oct 2026
Artificial Analysis: Time to first answer tokenseconds, lower is better 7.62 52 of 137 gpt-oss-20b (low reasoning) 20 Oct 2026
Artificial Analysis: τ-bench banking% of tasks, higher is better 7.0% 141 of 175 gpt-oss-20b (high reasoning) 20 Oct 2026
Artificial Analysis: τ²-bench telecom% of tasks, higher is better 60.2% 150 of 286 gpt-oss-20b (high reasoning) 20 Oct 2026
Artificial Analysis: τ²-bench telecom% of tasks, higher is better 50.3% 163 of 286 gpt-oss-20b (low reasoning) 20 Oct 2026

Not shown: Not your codebase or tools.

UGI Leaderboard · reported by UGI Leaderboard · 9 results
MeasureValueRankConfigurationDated
UGI: requested-length error% off the requested word count, lower is better 282.0% 379 of 379 gpt-oss-20b (high reasoning) 11 Oct 2025
UGI: requested-length error% off the requested word count, lower is better 98.0% 356 of 379 gpt-oss-20b (low reasoning) 6 Sep 2025
UGI: requested-length error% off the requested word count, lower is better 186.0% 375 of 379 gpt-oss-20b (medium reasoning) 7 Sep 2025
UGI: style adherencescore from 0 to 1, higher is better 0.27 375 of 379 gpt-oss-20b (high reasoning) 11 Oct 2025
UGI: style adherencescore from 0 to 1, higher is better 0.30 344 of 379 gpt-oss-20b (low reasoning) 6 Sep 2025
UGI: style adherencescore from 0 to 1, higher is better 0.31 311 of 379 gpt-oss-20b (medium reasoning) 7 Sep 2025
UGI: writing scorescore out of 100, higher is better 10.9 371 of 379 gpt-oss-20b (high reasoning) 11 Oct 2025
UGI: writing scorescore out of 100, higher is better 24.6 316 of 379 gpt-oss-20b (low reasoning) 6 Sep 2025
UGI: writing scorescore out of 100, higher is better 24.5 317 of 379 gpt-oss-20b (medium reasoning) 7 Sep 2025

Not shown: Not other format limits such as character counts or bullet counts.

Test gpt-oss-20b on your own prompt.

Benchmarks aren't your task. Run it next to cheaper and newer models on your prompt, with every answer checked and costed.